1. Field of the Invention
This invention relates to streaming data transmission (e.g. voice) over packet networks, and is particularly concerned with bandwidth-efficient and reliable DS-X transmission within distributed communication systems utilizing a packet network.
2. Description of the Related Art
Transferring streaming media, particularly voice and video, over a packet network presents challenges not faced with traditional time-switched systems. Traditional packet transport technologies and network/application protocols, such as TCP/IP FTP, and Ethernet are optimized for reliable data transfer, rather than the timeliness of the transfer. While such characteristics are great for transferring a data file or an application that will not work even if a small piece of data is corrupt or missing, connection and delivery confirmation delays make the transmission of time-sensitive streaming media less desirable.
Switched-packet Asynchronous Transfer Mode (ATM) does represent a convenient standard for transmitting voice, data, and video signals at very high speeds (25 Mbits/sec and higher) over packet networks. The increasing deployment of ATM-based networks, particularly on customer premises, has created a demand to provide high quality transport, switching, and call processing of voice communications over a packet-oriented network with minimal delay. The delay must not exceed that currently permitted for traditional centralized Private Branch Exchanges (PBXs), and the voice quality must be at least as good as such systems to warrant upgrading distributed ATM voice services.
A known distributed PBX architecture leveraging an ATM-based network is shown in FIG. 1. This PBX architecture includes several end node devices (e.g. access multiplexors 100, communications switch 120) distributed at the edges of a packet network transport. Access multiplexors 100 (access mux) are coupled to the ATM network (or single switch) 180 via links 150 carrying ATM cell traffic at STS-1/OC-1 speeds or higher. These muxes 100 convert locally-originated, digital channels representing e.g. PCM-encoded voice/data traffic at DS-X speeds (i.e. the US plesiochronous digital hierarchy DS-0, DS-1, DS-2, DS-3, DS-4, etc. or the European E-1, E-2, . . . , hereinafter referred to as xe2x80x9cDS-X trafficxe2x80x9d) into packets, receive and reconstruct remotely-originated voice/data packets back into native DS-X form, and route each to their intended destinations. High speed link 170 (e.g. SONET STS-3c, OC-12) couples the communications switch 120 to the packet network 180. Logical or virtual connections, known in ATM parlance as virtual channels, are established between the access muxes 100 and the communications switch 120 within the packet network 180 to permit transfer, call processing and switching of the packetized DS-X traffic.
As with the case with traditional centralized PBX systems, synchronous Access Devices (ADs) 110, 115 and 118 are coupled to these access muxes 100 to collect voice/data information (such as a 64 Kbps DS-0 voice channel) originating from telephony devices such as digital/analog extensions 130/132 or videophone 134, as well as T1/E1 trunks 136. Once collected using known time-division multiplexing techniques, the ADs relay synchronous DS-X traffic to the access mux 100 using signal lines 120. Of course, ADs 110, 115, 118 are also used to distribute outgoing DS-X traffic extracted by the access mux 100 over the signal lines 120 to the appropriate destination telephony device or trunk.
In order to leverage the ATM network for voice communications, each access mux 100 converts DS-X traffic into ATM cells. According to the ATM standard, each cell is a fixed fifty-three bytes or octets long including a forty-eight byte payload and a five byte header. For outgoing calls, each access mux 100 packs the DS-X traffic into the 48-byte payload using conventional methods for multiplexing voice DS0s into ATM virtual channels, such as T1/E1 Circuit Emulation Service (CES) based on ATM Adaptation Layer 1(AAL 1) or other ATM adaptation layers for carrying voice. For incoming calls, each access mux determines the destination of the DS-X traffic encapsulated within each received cell based in part on the virtual channel or path information contained in the cell as well as its location within the cell payload. The access mux 100 then converts the packets to the carrier format for the signal line(s) 120 corresponding to the intended destination device.
However, without further refinement, this architecture is prone to cell delay (especially where the customer""s packet network is complex or geographically diverse), which can seriously degrade the quality of service for voice and streaming data it is slated to carry. To minimize cell-fill delay here, which can form a major component of the overall architecture delay, it has been proposed that each DS-0 channel serviceable by the access mux 100 contributes one byte to the ATM cell(s) and the so-filled cells are transported within a 125 xcexcsec frame, corresponding to the well-established 8 KHz sampling rate for voice. To reduce communication switching setup, complexity and delay, as well as simplify cell composition/decomposition logic within each access mux 100, each available DS-0 is statically assigned to a unique octet slot within one of the cells filled and transported within the 125 xcexcsec frame. To further simplify network complexity, along with minimizing dynamic connection setup and tear down delays associated with ATM switched virtual channels (SVCs), permanent virtual channels (PVCs) are established within the packet network 180 at system configuration through management action to transport the DS-0 packed cells between the access mux 100 and the switch 120.
In applying these techniques to the architecture of FIG. 1, a number of Constant Bit Rate (CBR) PVCs capable of bearing the simultaneous DS-0 capacity of each access mux 100 is established within the packet network 180. A PVC will be used to bear one ATM cell per 125 xcexcsec frame, and cell payload will contain a byte of data from each of 48 DS-0 channels statically mapped into its 48 octet slots. Since the placement of the DS-0 byte within a particular PVC and octet slot identifies the DS-0 connection (originating/terminating device), all cells must be broadcast within a given 125 xcexcsec frame, whether or not traffic actually exists on the connection. Typically, cell octet slots corresponding to DS-0 channels not bearing activity within a given frame are stuffed with pre-established xe2x80x9cdon""t carexe2x80x9d data.
In the above-stated combination, these design choices seriously impact bandwidth efficiency. For example, if the access muxes 100 could simultaneously service sixteen (16) ADs, each defining thirty-two (32) DS-0 channels (which is typical for such units), a minimum of 512 DS-0 channel bytes would accordingly need to be transferred every 125 xcexcsec frame, whether or not traffic actually exists on those channels. Eleven PVCs would be required to fully transport these bytes with padding left over, and accordingly eleven ATM cells would need to be transmitted every frame. FIG. 3 graphically indicates the DS0 mapping required for the 16 32 DS-0 channel ADs plotted against the payload portions of the cells 330-380 to be filled, with xe2x80x9cXxe2x80x9d representing 16 empty octet slots at the end of the last cell 380 which are padded out in order to complete the payload. Thus, in this example, regardless of the actual DS-0 loading (meaning the number of DS-0 channels actually allocated or reserved for use) or active DS-0 traffic experienced, approximately 37.3 Mbits/sec of bandwidth would be required, exclusive of any control signaling. In view of the fact that this represents almost a quarter of the available bandwidth of an STS-3c link, and that the access mux 100 almost never experiences full loading (and is actually quite sparse in the typical PBX application), this configuration appears to be an undesirable choice where the packet network 180 is not devoted exclusively for PBX transport.
Such inefficiencies are carried over and magnified at the communications switch 120. According to the architecture of FIG. 1, communications switch 120 will be responsible for handling switching and call management responsibilities for the packetized DS-X traffic generated by and destined for each access mux 100. Assuming that each access mux 100 contributes 37.3 Mbits/sec of bandwidth, the link 170 must be able to bear A*37.3 Mbits/sec, exclusive of control and management signals, where A represents the number of access muxes 100 serviced by the communication switch 120. Thus, link 170 must typically support a magnitude higher simultaneous bandwidth than links 150 (e.g. OC-12 vs. STS-3c), which adds to the cost and complexity to communications switch 120.
Moreover, it should be realized that the enormous number of traffic bearing cells entering the communications switch 120 packed according to the above-described techniques and criteria require commensurately enormous and expensive computing resources within the switch 120 if switching and call management responsibilities are to be completed within the required 125 xcexcsec frame.
Yet another problem arises from the use of PVCS under the current ATM standards. PVCs used to bear the packetized DS-X traffic must be established and maintained by the PBX-specific components (either the access muxes 100 or the communications switch 120), despite the limited ongoing control and management they can exert over the packet network 180. This, coupled with the aforementioned static mapping criteria, makes it difficult for the PBX of FIG. 1 to recover from a lost PVC without dropping all connections and reinitializing.
Therefore, it would be desirable to develop an improved distributed PBX architecture which more efficiently utilizes available packet network bandwidth without materially sacrificing cell delay performance. It would also be desirable to develop a distributed PBX architecture which is more tolerant of connection failures within the packet network.
In accordance with these and related desires, a novel and nonobvious method for transporting DS-X traffic over a packet network is proposed. Specifically, according to the present invention, virtual connection or slot provisioning and/or cell concentration techniques are used to compact the amount of DS-X traffic broadcast between communications system devices such as the end node and the communications switch to spare bandwidth. In the case of provisioning, a configured DS-X loading of an end node supporting DS-X traffic is ascertained. This configured DS-X loading sets forth the actual loading in terms of digital channels being simultaneously supported and/or number of number of access devices coupled to and being serviced by the end node. In turn, preferably a minimum number of virtual connections for bearing packetized DS-X traffic are established, either at communications system configuration or as needed.
Through provisioning, only the virtual connections needed to support all the digital channels in actual service or the DS-X capacity of all actually-serviced access devices need be established, and packetized DS-X traffic need only flow on those so-established virtual connections. By contrast, conventional virtual connection establishment methods teach that the full capacity of the end node define how many virtual connections to establish and maintain. Thus, through provisioning according to the present invention, bandwidth is spared where the configured capacity of the end node is less than its full capacity such that less total virtual connections are needed to bear the DS-X traffic, even where static slot mapping is utilized.
Compaction of the DS-X traffic bearing cells themselves may be accomplished through concentration processing according to the present invention. In such concentration processing, a dynamic association between the virtual connections and the DS-X traffic is established, typically on a per frame basis. According to this invention, only those digital channels entering the node within a given time frame which actually bear DS-X traffic will be assigned a slot within packet. A DS-X trafficxe2x80x94packet correspondence, such as a slot map, may be compiled and updated as necessary based on the active DS-0 channels within the given frame, and appropriate slot assignment data, such as the updated slot map or any changed assignment data may be exchanged between communication system devices in order to properly compose and decompose the DS-traffic packed in the packets and preserve intended channel connections therebetween.
According to the present invention, concentration processing may be augmented by provision processing to further reduce bandwidth requirements for transporting the packetized DS-X traffic.
In other aspects of the present invention, permanent virtual channels or soft permanent virtual connections established through Private Network-Network Interface management and signalling may be used for the virtual connections where ATM is chosen as the transport protocol. Soft permanent virtual connections are seen as being especially advantageous since their ongoing management and re-routing is performed by the ATM packet network being leveraged by the communications system, rather than the communications system or its components itself.
Also, the present invention contemplates an end node apparatus incorporating these provisioning, concentration processing, and/or virtual connection establishment features to reduce the bandwidth requirements of packetized DS-X traffic flowing into or out of the node.
Other aspects and features of the present invention will become apparent to those ordinarily skilled in the art upon review of the following description of the specific preferred embodiments of the invention in conjunction with the accompanying figures.